[Debichem-devel] Bug#834786: Bug#834786: gromacs: FTBFS on hurd-i386 and m68k: test suite errors

[Nicholas Breen]

On Thu, Aug 18, 2016 at 11:35:07PM -0400, Aaron M. Ucko wrote:
> Source: gromacs
> Version: 2016-1
> Severity: important
> Justification: fails to build from source (but built successfully in the past)
> 
> Builds of gromacs for hurd-i386 and m68k each failed with (unrelated)
> test suite errors.  Could you please take a look?
> 
> hurd-i386:
> 
>   /«PKGBUILDDIR»/src/gromacs/hardware/tests/hardwaretopology.cpp:169: Failure
>   Expected: (n.memory) > (0), actual: 0 vs 0
>   [  FAILED  ] HardwareTopologyTest.NumaCacheSelfconsistency (0 ms)

I'm already logged into the hurd-i386 porterbox to work on this, in fact.  At
the moment I suspect it's simply a lack of Hurd feature support in hwloc, and
the test will need to be disabled; lstopo output looks fine.

> m68k:
> 
>   /<<PKGBUILDDIR>>/src/testutils/refdata.cpp:817: Failure In item:
>   /NormalDistribution/[0] Actual: -1.1957091691864972 Reference:
>   -1.1957091691864985 Difference: 1.33227e-15 (6 double-prec. ULPs, rel.
>   1.33e-15) Tolerance: abs. 8.88178e-16, 4 ULPs (rel. 8.88e-16) [  FAILED  ]
>   NormalDistributionTest.Output (10 ms)

This one's a little different.  It's directly caused by a packaging bug (one of
the sub-builds doesn't obey DEB_BUILD_OPTIONS=nocheck, which is set by default
on m68k), and I'll fix that, but I'd be happier if it *could* pass all the
tests....



-- 
Nicholas Breen
nbreen at debian.org